Production of Unilamellar Vesicles Using an

We use water-in-oil emulsions stabilized by the material for the inner leaflet and ... (3) Barron, L. G.; Meyer, K. B.; Szoka, F. C. J. Hum. ... was p...
4 downloads 0 Views 245KB Size
2870

Langmuir 2003, 19, 2870-2879

Production of Unilamellar Vesicles Using an Inverted Emulsion Sophie Pautot,† Barbara J. Frisken,‡ and D. A. Weitz*,† Department of Physics and DEAS, Harvard University, Cambridge, Massachusetts 02138 and Department of Physics, Simon Fraser University, Burnaby, BC V5A 1S6 Received June 20, 2002. In Final Form: January 2, 2003 We investigate a method for the controlled assembly of unilamellar vesicles consisting of bilayers assembled one leaflet at a time. We use water-in-oil emulsions stabilized by the material for the inner leaflet and produce vesicles by passing the water droplets through a second oil-water interface, where they become coated with the outer leaflet. We have used this technique to form vesicles from lipids, mixed lipid and surfactant systems, and diblock copolymers. The stability of lipid-stabilized emulsions limits the range of sizes that can be produced and the vesicle yield; nevertheless, there are several advantages with this emulsion-based technique: It is possible to make unilamellar vesicles with sizes ranging from 100 nm to 1 µm. Moreover, the process allows for efficient encapsulation and ensures that the contents of the vesicles remain isolated from the continuous aqueous phase. To illustrate possible applications of this technique, we demonstrate the use of vesicles as microreactors where we polymerize actin through the addition of magnesium and show that the polymerization kinetics are unaffected by the encapsulation.

Introduction Bilayers of lipid molecules that fully enclose a fixed volume of fluid are called vesicles or liposomes. They are ubiquitous in cells where they encapsulate and isolate intracellular fluids and often contain specific proteins or other macromolecules. Vesicles are also used to transport macromolecules through the blood stream or through the skin, leading to the widespread use of vesicles in cosmetics and drug delivery.1-3 In addition, they serve as a model system for the study of the fundamental properties of lipid bilayer membranes. Ideally, these applications require simultaneous control of vesicle size, unilamellarity, encapsulation yield, biocompatibility, and lipid composition, but none of the vesicle production techniques currently used fulfills every requirement. Sonication can be used to produced vesicles with radii smaller than 50 nm (small unilamellar vesicles or SUV) but this produces a highly heterogeneous size distribution.4-6 Alternatively, large multilamellar vesicles (MLV), formed spontaneously in excess water, can be extruded through a narrow pore size filter to produce a more monodisperse distribution of large unilamellar vesicles (LUV).7 By contrast, production of giant unilamellar vesicles (GUV) that range in size from 1 to 20 µm is usually accomplished by rehydration or swelling8 thin lipid films with water or by electroformation;9 the resultant suspension contains a broad distribution of sizes, and only a few of the vesicles produced are † ‡

Harvard University. Simon Fraser University.

(1) Gregoriadis, G. FEBS Lett. 1971, 14, 95. (2) Nicolau, C.; Paraf, A. Liposomes, drugs and Immunocompetent Cell Functions; Academic Press: New York, 1981. (3) Barron, L. G.; Meyer, K. B.; Szoka, F. C. J. Hum. Gene Therapy 1998, 9, 315-323. (4) Gregoriadis, G. Preparation of Liposomes; CRC Press: Boca Raton, FL, 1984; Vol. 1. (5) Ostro, M. J. Liposomes; Marcel Dekker Inc.: New York, 1983. (6) Philippot, J. R.; Schubert, F. Liposomes as Tools in Basic Research and Industry; CRC Press: Boca Raton, FL, 1995. (7) Olson, F.; Hunt, C. A.; Szoka, F. C.; Vail, W. J.; Papahadjopoulos, D. Biochim. Biophys. Acta 1979, 9, 557. (8) Lasic, D. D.; Yechezkel, B. Handbook of Non Medical Applications of Liposomes; CRC Press: Boca Raton, FL, 1996. (9) Angelova, M. I.; Dimitrov, D. S. Faraday Discuss. 1986, 81, 303.

unilamellar. Vesicles produced by all of these techniques are formed in the same medium that they encapsulate, which limits their utility for encapsulation of precious drugs or other active agents available in limited quantities. One technique, called reverse evaporation,10 overcomes this deficiency but depends on the use of organic solvents, trace amounts of which remain in the bilayer and can contaminate the internal aqueous phase. Finally, to successfully produce vesicles, each technique has specific requirements for either the lipids or the buffer used. This lack of flexibility has negatively impacted the use of vesicles for drug delivery; their role is typically limited to that of simple carrier rather than the dynamic and functional compartments they could become. To help overcome some of these limitations, we have investigated a method of vesicle production that involves the assembly of two independently formed monolayers of amphiphilic molecules into unilamellar vesicles. The first step is the preparation of a stable inverted emulsion, as shown in Figure 1A. The aqueous solution to be emulsified is mixed with the oil phase containing surfactant molecules. As the water drops are formed, the surfactant molecules dispersed in the oil adsorb at the surface of the drops to form a monolayer that helps stabilize the emulsion, reducing coalescence of the drops. This monolayer will form the inner leaflet of the bilayer that will make up the final vesicle bilayer. We then pour a second oil phase, also containing surfactant molecules, over the top of an aqueous phase, as shown in Figure 1B. Because the oil is less dense, it remains above the aqueous phase, and surfactant molecules dispersed in the oil phase diffuse from the bulk to the water/oil interface to form a monolayer. After the interface is fully covered, we gently pour the inverted emulsion on top of the oil-buffer interface as shown in Figure 1C. The water droplets in the inverted emulsion are heavier than the oil phase and sediment toward the oil-buffer interface. Upon crossing the interface, the emulsion droplets pick up a second layer of lipid, forming a bilayer and transforming the emulsion (10) Szoka, F.; Papahadjopoulos, D. Proc. Natl. Acad. Sci. 1978, 75, 4194-4198.

10.1021/la026100v CCC: $25.00 © 2003 American Chemical Society Published on Web 02/19/2003

Production of Unilamellar Vesicles

Figure 1. Schematic illustration of the synthesis of vesicles from an inverted emulsion. (A) Water is emulsified in oil with lipid as the surfactant, forming a stable inverted emulsion. (B) The water that will receive the final vesicles is placed in a second vial, and lipid-saturated oil is poured on top of it. A lipid monolayer forms at the oil-water interface. (C) Preparation A is then gently poured on top of preparation B. The inverted emulsion droplets are heavier than the oil that contains them and sediment into the second water phase. As they pass through the interface where the second layer of lipid is sitting, the bilayer is completed and the final vesicles are formed.

droplets into unilamellar vesicles. After sedimentation, the vesicles are in the bottom aqueous buffer. An emulsion-based method of vesicle production has several obvious advantages. During the entire process, the encapsulated material remains distinct from the hosting buffer. The vesicles are thus filled with the solute of interest directly during their formation, making high encapsulation efficiency possible. The use of emulsification for encapsulation in the first step makes it independent of the type of amphiphilic molecule used, of the size or charge of the macromolecules to be encapsulated, or of the ionic strength of the buffer. It improves the fabrication of unilamellar vesicles by making a wider range of sizes accessible with a single technique. It provides a means to control the composition of each leaflet making it possible to use a broader choice of amphiphilic molecules to assemble the bilayer. It also makes it possible to form asymmetric bilayers. Because every component of the vesicle can be controlled, this technique provides the potential for new applications in biomaterials engineering including the possibility of transforming these vesicles into microreactors where reactive molecules can be efficiently encapsulated and isolated from the bulk and where remotely triggered reactions can occur. This method has been discussed before but has not been implemented widely. It was demonstrated as a method of vesicle production using a benzene:water:egg-PC emulsion, and the possibility of making asymmetric vesicles was proposed.11 More recently, an encapsulation efficiency (11) Trauble, H.; Grell, E. Neurosci. Res. Programm Bull. 1971, 9, 373-380.

Langmuir, Vol. 19, No. 7, 2003 2871

of 60% was reported using vesicles produced from a decane: water:egg-PC emulsion.12 In both cases, sonication was used to prepare the initial emulsion and electron microscopy was used to determine the vesicle sizes; they were consistently less than 200 nm in diameter. In this paper, we present a detailed description of the inverted-emulsion method and we discuss its limitations. We show that, by application of suitable emulsification techniques, it is possible to make vesicles ranging in size from 0.1 µm to 1 µm with a variety of surfactant-like molecules. We demonstrate that the vesicles are unilamellar and that it is possible to achieve encapsulation efficiencies as high as 98%. In other work,13 we show that it is possible to use this technique to make asymmetric vesicles. We also discuss limitations to this technique that occur when lipids are used to stabilize the emulsion; these include control of the size and number of vesicles formed. We have observed that lipids do not behave like most surfactants. Thus, the emulsification process is sometimes not fully controlled and the slow equilibration of lipid monolayers at oil-water interfaces limits replenishment of this interface and thus vesicle yield. We discuss the behavior of lipids in an oil-water environment and how this impacts the control of the size of the emulsion, the composition of each leaflet of the bilayer, and ultimately the number of vesicles produced. Finally, to illustrate the potential of this technique, we demonstrate the encapsulation of a variety of molecules without loss of their activity and show that it is possible to use vesicles as microreactors. Materials and Methods Materials. The lipids used included egg-PC (egg-phosphatidylcholine), POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine), POPS (1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine), and DOPS (dioleoyl-sn-glycero-3-phospho-L-serine). We also used fluorescently labeled lipids including NBD-PC (1-palmitoyl-2(6-((7-nitro2-1,3-benzoxadiazol-4-yl) amino) caproyl)-sn-glycero3-phosphocholine) and NBD-PS (1-palmitoyl-2-(6-((7-nitro2-1,3benzoxadiazol-4-yl) amino) caproyl)-sn-glycero-3-(phospho-Lserine)) and rhodamine-DHPE (1,2-dihexadecanoyl-sn-glycero3-phosphoethanolamine). All of these lipids were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL) as 99% pure chloroform stock solution and were used without further purification. Biotinconjugated lipid (N-((6-biotinoyl)amino)hexanoyl)-1,2-dihexadecanoyl-sn-glycero-3-phosphoethan olamine) was obtained from Molecular Probes (Eugene, OR). A diblock copolymer PABu-PAM provided by Rhodia (Cranbury, NJ) was used to prepare polymersomes. This is a symmetric diblock copolymer composed of a polybutylacrilate block and a polyacrilamide block, both with Mw ) 8000 g/mol. The encapsulation of macromolecules was demonstrated with a 1 wt % solution of dextran, Mw)10 000 g/mol, tagged with Texas Red (Molecular Probes, Eugene, OR). The encapsulated solution used to probe the transport of divalent cations across the bilayer was a polymer solution prepared by suspending 10 000 Mw dextran tagged with Oregon Green 488 bapta-1, a calcium sensitive dye (Molecular Probes, Eugene, OR), in an aqueous buffer consisting of 0.1 M KCl with 10 mM Tris at pH ) 8. The solution was dialyzed overnight against a buffer free of calcium and magnesium to remove divalent cations that might be present in the polymer powder; the final concentration was then set at 1 mg/mL. Finally, to study the polymerization of actin inside the vesicles, we used actin from acanthamoeba purified according to the protocol established by Gordon et al.14 We prepared N-(1pyrenyl) iodoacetamide-labeled F-actin15 and used 20% labeled (12) Zhang, L.; Hu, J.; Lu, Z. J. Colloid Interface Sci. 1997, 190, 76-80. (13) Pautot, S.; Frisken, B. J.; Weitz, D. A. Unpublished. (14) Gordon, D. J.; Eisenberg, E.; Korn, E. D. J. Biol. Chem. 1976, 251, 4778-4786. (15) Kouyama, T.; Mihashi, K. Euro. J. Biochem. 1981, 114, 33-38.

2872

Langmuir, Vol. 19, No. 7, 2003

Pautot et al.

actin and 80% nonlabeled actin to study the kinetics of polymerization of the actin upon addition of magnesium. Preparation of the Vesicles. The choice of the organic phase used to prepare the inverted emulsion is dictated by the balance between the solubility of the lipid in oil, the stability of the inverted emulsion, and the possible presence of oil in the final bilayer. Unlike previous attempts to use similar techniques,12,16,17 we use a sufficiently long-chain alkane to ensure that the alkane molecules are not incorporated into the final bilayer18 and that a stable, inverted macroemulsion is formed. We observed that the use of alkanes with chains shorter than 10 carbons resulted in a microemulsion where we were no longer able to control the size of the water droplets. We found that dodecane, a 12 carbon alkane, was most suitable for our applications, but a variety of oils can be used. Alkanes with chain lengths ranging from 12 to 17 carbons resulted in stable macroemulsions, as was squallene, a complex carbon alkane chain of high molecular weight. We have checked the composition of the bilayer by making a thinlayer chromatography analysis of the final vesicles and comparing this to that of pure lipids and pure alkane samples. We could not detect the presence of alkane indicating that, if alkane molecules were present, they made up less than 5% of the bilayer. This is the sensitivity limit of this technique; other techniques were not pursued. The best alkane for an oil-free bilayer is squallene, which has the advantage that it is immiscible in lipid bilayers18 while still providing a good continuous phase for emulsion preparation. A wide variety of surfactants can be used to stabilize an inverted emulsion. In this work, we used nonionic surfactants, lipids, and diblock copolymers, chosen according to geometrical arguments19 such that

( )

v g1 (a0lc)

(1)

where v is the volume of the hydrocarbon chain, a0 is the surface area of the polar head, and lc is the chain length. When this ratio is greater than one, the area of the hydrophobic part of the molecule is larger than that of the hydrophilic part, which favors the stabilization of an inverted emulsion. Lipids have a ratio only slightly higher or lower than one depending on the polar headgroup and salt conditions; this fact makes them a poor surfactant with which to stabilize an emulsion. The inverted emulsion was usually prepared by first dispersing the surfactant in the oil phase. For multicomponent lipid mixtures, the lipids were first dissolved together in chloroform to obtain a homogeneous dispersion. The chloroform was then evaporated under a stream of nitrogen to form a dry thin film on the surface of a glass vial, which swelled easily when dodecane was added. To fully disperse the lipid molecules in dodecane, the suspension was placed in a sonication bath overnight at 25 °C. Surfactants in liquid form, like Span80, a nonionic surfactant, were added directly to the dodecane. Finally, for samples prepared with the diblock copolymer PABuPAM (Rhodia, Cranburry, NJ), it was more convenient to dissolve the powder directly in the aqueous solution at 0.1 wt % to prepare the inverted emulsion. The emulsification of the aqueous solution and surfactantsaturated dodecane can be achieved by various methods.20 The amount of shear and the method by which it is applied to the suspension are responsible for the mean size and polydispersity of the emulsion produced. Four different emulsification techniques were used: shear by gentle stirring, shear using a mixer, extrusion, and sonication. For lipids and nonionic surfactants, a crude inverted emulsion was prepared either by vortexing the solution for several minutes or by gently stirring the suspension with a magnetic stir bar for 1-3 h. The emulsion produced was (16) Szoka, F. C.; Papahadjopoulos, D. Annu. Rev. Biophys. Bioeng. 1978, 9, 467-508. (17) Xiao, Z. D.; Huang, N. P.; Xu, M. H.; Lu, Z.; Wei, Y. Chem. Lett. 1998, 3, 225. (18) McIntosh, T. J.; Simon, S. A.; MacDonald, R. C. Biochim. Biophys. Acta 1980, 597, 445-463. (19) Israelachvili, J. Intermolecular and Surface Forces; Academic Press: London, 1992. (20) Hunter, R. J. Foundations of Colloid Science, Volume II; Oxford Science Publications: New York, 1989.

very polydisperse with a mean size around 1 µm. More controlled preparation was achieved by using a shear mixer with variable rotation speed (Ultra-Turrax T25, IKA, Wilmington, NC) or extrusion of the suspension through a narrow pore size polycarbonate filter (Poretics Corp., Livermore, CA). Extrusion was also used to reduce the polydispersity and the size of the droplets in an emulsion produced by other techniques. Further purification steps are then required to extract the inverted emulsion.21 Finally, in some cases, such as the diblock copolymer, the best results were obtained by sonicating the suspension. Unless otherwise specified, the emulsified aqueous solution consisted of 100 mM NaCl and 5 mM Tris buffered at pH ) 7.4. The volume fraction of the aqueous solution was 0.5% of the total volume. To prepare for the final step in the assembly, we placed 3 mL of aqueous buffer in a centrifuge tube (Falcon tube, 50 mL with a 1-in. diameter) and then gently poured about 2 mL of dodecane containing amphiphilic molecules for the outer monolayer on top of the buffer. Some of these molecules diffused to cover the oil-water interface. For most surfactants, adsorption at an interface is diffusion limited22 and full coverage should occur in only a few minutes. The adsorption mechanism for lipids appears to be more complex; we found that it takes from 30 min for charged lipids up to 90 min for zwitterionic ones to achieve full coverage.23 Once the interface was fully covered, the inverted emulsion was added to the Falcon tube. The volume of the emulsion added was varied from 0.1 mL to 1 mL to ensure that the total area of the emulsion did not exceed the total area of the monolayer available to supply the outer leaflet. After the three layers were assembled, the tube was placed in a spin-bucket centrifuge and spun at 120 g for 5-10 min. Since water is denser than oil, the water droplets sediment toward the oil-buffer interface. Centrifugation was used to enhance the rate of sedimentation, decreasing the time required to transfer the droplets from several hours to a few minutes. Characterization. (1) Microscopy. All of our real-space observations were done using an inverted microscope (Leica) used either in normal or confocal mode. Some of the pictures were obtained with a 100x phase contrast oil immersion objective while the fluorescence pictures were obtained with a 100x DIC oil immersion objective. Since freely diffusing vesicles smaller than 1 µm move in and out of the focus plane rapidly, it is difficult to image them and measure their size by microscopy. To overcome this, we used streptavidin/biotin chemistry24 to fix the vesicles to the coverslip, confining them to a plane and improving the image. For this purpose, vesicles were prepared with 1% biotin-conjugated lipid. After the vesicles were made, they were first incubated with a 1:1 molar ratio of free streptavidin (Sigma, St. Louis, MO) for 15 min, resulting in streptavidin-coated vesicles, and were then incubated for 10 min on a cover glass previously coated with biotinylated albumin (Sigma, St. Louis, MO). The cover glass was washed five times with buffer to eliminate unbound vesicles. The bound vesicles did not seem to adhere to the coated cover glass with a wide contact area; although attached to the surface, they still moved about their anchoring point, suggesting that the binding was relatively weak. (2) Fluorescence Experiments. Fluorescence measurements were performed on a luminescence spectrometer (Perkin-Elmer LS 50B). The instrument was used to monitor changes in the intensity of emitted light of wavelength λem for a fixed excitation wavelength λexc while the local environment of the dye was modified. Experimental parameters such as excitation and emission wavelength were determined by measuring the excitation and the emission spectrum for each sample. We used a fluorescence quenching assay25,26 to measure the distribution of tagged lipids between inner and outer monolayers. (21) Bibette, J. J. Colloid Interface Sci. 1991, 147, 474-478. (22) Ferri, J. K.; Stebe, K. J. Adv. Colloid Interface Sci. 2000, 85, 61-97. (23) Pautot, S.; Frisken, B. J.; Cheng, J.-X.; Xie, S.; Weitz, D. A. Unpublished. (24) Bayer, E. A.; Wilchek, M. Methods Biochem. Anal. 1980, 26, 1-45. (25) McIntyre, J. C.; Sleight, R. G. Biochemistry 1991, 30, 1181911827. (26) Gruber H. J.; Schindler, S. H. Biochim. Biophys. Acta 1994, 1189, 212-224.

Production of Unilamellar Vesicles

Langmuir, Vol. 19, No. 7, 2003 2873

A suspension of vesicles was prepared with 0.5% of a fluorescently labeled lipid. The fluorescence of the vesicle solution was measured before and after addition of a quenching solution of sodium hydrosulfite (1 M Na2S2O4 in 5 mM TES at pH ) 9, prepared daily). When in the vicinity of the fluorophore, the sodium hydrosulfite extinguishes the fluorescence by reducing the dye. As this molecule does not diffuse across the lipid bilayer, the addition of the quencher to the vesicle suspension results in the extinction of only the dye located on the outer leaflet of the bilayer. The dye on the inner leaflet was exposed by adding detergent (Triton X-100 reduced) to lyse the bilayer. The excitation wavelength for these measurements was set at λexc ) 470 nm, and the emission of fluorescence was measured at λem ) 550 nm. Finally, fluorescence measurements were used to probe the polymerization of pyrene-labeled actin protomers. Polymerization of actin containing 20% actin protomer labeled with pyrene leads to an increase in fluorescence intensity of the pyrene.15 This increase in fluorescence is due to a modification of the pyrene environment when the actin protomers associate during the polymerization. Furthermore, the spectral signature of pyrene F-actin is slightly different from pyrene G-actin allowing the polymerization to be monitored. The excitation wavelength used was λexc ) 365 nm, and the change in the emitted fluorescence intensity was monitored at λem ) 407 nm after the polymerization was initiated by the addition of salt. (3) Dynamic Light Scattering. The apparatus used for the lightscattering experiments was an ALV DLS/SLS-5000 spectrometer/ goniometer (ALV-Laser GmbH, Langen, Germany). The light source for the experiments was an argon ion laser of wavelength 514.5 nm (Coherent, CA). Light scattered by the sample was detected at an angle of 90° from the transmitted beam, where the effects of reflection are minimized. The sample cell was placed in a toluene bath maintained at a temperature of 25 °C. The size and size distribution of the inverted emulsion and the vesicles were characterized by dynamic light-scattering (DLS) measurements. In DLS, we measure the normalized time autocorrelation of the intensity of the scattered light g(2)(τ), which is related to the time autocorrelation function of the scattered electric field light g(1)(τ).27 For a dilute suspension of monodisperse particles, g(1)(τ) decays exponentially with a decay rate Γ ) Dq2, where q is the magnitude of the scattering wave vector and D is the diffusion coefficient, which can be related to the hydrodynamic radius though the Stokes-Einstein relation. For a polydisperse sample, g(1)(τ) is no longer a single exponential because particles with different sizes have different diffusion coefficients and hence different decay rates. The effect of a distribution of decay rates G(Γ) on g(1)(τ) is given by

g(1)(τ) )

∫ G(Γ)e ∞

0

-Γτ



(2)

A regularized fit routine supplied by the ALV instrument, which is similar to CONTIN 2DP,28,29 was used to obtain G(Γ) when an expansion in term of cumulants30 failed to fit the data because the size distribution was multimodal. The resulting distributions have to be viewed with considerable caution, and they are used here only to estimate the populations present in our samples when other types of analysis are not possible.

Vesicle Production Characterization. (1) Visualization. The vesicles produced with this technique can be visualized by microscopy. Figure 2A shows an image, obtained with phase contrast microscopy, of polydisperse vesicles prepared from an inverted emulsion. The emulsion was prepared by gently stirring for 3 h 1 vol % aqueous buffer in dodecane containing 0.05 mg/mL POPC. The resulting emulsion was polydisperse with droplet sizes ranging from (27) Berne, B. J.; Pecora, R. Dynamic Light Scattering; John Wiley and Sons: New York, 1976. (28) Provencher, S. W. Comput. Phys. Commun. 1982, 27, 213-227. (29) Provencher, S. W. Comput. Phys. Commun. 1982, 27, 229-242. (30) Koppel, D. E. J. Phys. Chem. 1972, 52, 4814-4820.

Figure 2. (A) Phase contrast microscopy image of larger, polydisperse POPC vesicles assembled using an inverted emulsion prepared with a low-shear technique. (B) Fluorescence microscopy images of egg-PC vesicles formed from an emulsion extruded through a 1-µm filter.

1 to 4 µm in diameter. Figure 2B shows a suspension of monodisperse egg-PC vesicles obtained from an emulsion that was extruded several times through a 1-µm pore filter. One mole % of rhodamine was mixed with the egg-PC to allow visualization by fluorescence microscopy, and 1 mol % of biotin-conjugated lipid was added so that the vesicles could be bound to a streptavidin substrate. The emulsion used in this preparation was composed of drops about 500 nm in diameter, as well as some much larger drops of aqueous phase that was not fully emulsified, which were not used in the vesicle preparation. The vesicles that could be made most easily were relatively small in size, less than ∼500 nm in diameter; we found that the size distribution of these smaller vesicles was determined by that of the initial inverted emulsion. To illustrate this, we compare the autocorrelation functions measured for a vesicle sample with that of the inverted emulsion used to prepare it in Figure 3A. The inverted emulsion, stabilized with POPC and prepared by gentle stirring, was used to form vesicles after the larger drops had been removed by sedimentation. After the time scale for the emulsion data has been scaled by the ratio of the viscosities, the autocorrelation function for the inverted emulsion (triangles) is virtually indistinguishable from that of the vesicles (circles). The distributions of decay times are shown in Figure 3B. The distribution for the inverted emulsion is slightly broader, perhaps because of the presence of lipid aggregates in the sample. The mean of the two distributions is comparable; the mean radius of the emulsion distribution is 220 nm where that of the vesicle distribution is 170 nm. These results confirm that the size of the inverted emulsion dictates the size of the vesicles for these small droplets. (2) Determination of Vesicle Unilamellarity. We used a fluorescence quenching assay to determine the lamellarity of the vesicles. Figure 4 shows the time-dependence of the fluorescence intensity of POPS:NBD-PS vesicles made from an emulsion prepared by gentle stirring. At the first arrow, 60 µL of quencher solution was added. As the quencher diffuses to the surface of the vesicles, the intensity decreases and finally reaches a plateau when the fluorescence of all the NBD-PS lipids present in the outer leaflet of the bilayer has been extinguished. At the second arrow, Triton-X-100 was added. As the vesicles

2874

Langmuir, Vol. 19, No. 7, 2003

Figure 3. Comparison of the distribution of sizes between the emulsion and vesicles when the inverted emulsion contains small (1 µm) droplets. (A) DLS data taken at 90° for an inverted emulsion stabilized by POPS rescaled to take into account the difference in viscosity (triangles) and for the suspension of vesicles formed from this emulsion (circles). (B) Corresponding distribution of decay times for both samples. We observe differences between the distributions found for the emulsion (triangles) and that found for the vesicles (circles), particularly the loss of part of the largest inverted emulsion droplets upon sedimentation.

Alternatively, a direct scheme of droplet production33-35 may be used to restrict the formation of the emulsion drops to the desired size. For example, emulsions formed by extrusion can be used to produce large vesicles with relatively narrow size distributions, as we will show below. (4) Transfer of the Emulsion Droplets to the Aqueous Phase. The outer monolayer is formed as the emulsion droplets pass through the interface between the intermediate and aqueous phases. Successful completion of this step depends on stability of the emulsion and coverage of the interface layer. As shown in Figure 3, we are able to obtain good transfer of small droplets, less than ∼500 nm diameter. Larger droplets present a much greater problem when they are transferred to the final aqueous phase. Figure 8 shows data for an inverted emulsion stabilized by POPS and prepared by gentle stirring and a solution of vesicles formed from it without fractionation. Figure 8A shows the autocorrelation function for the inverted emulsion (triangles) rescaled to take into account the difference in viscosity and the autocorrelation for the vesicles produced (circles). The two curves depart from each other at large correlation times indicating that larger drops are present in the emulsion. The corresponding decay time distributions are shown in Figure 8B; the two curves confirm that the population of large objects is much less important in (34) Umbanhowar, P. B.; Prasad, V.; Weitz, D. A. Langmuir 2000, 16, 347-351. (35) Thorsen, T.; Roberts, R.; Arnold, F.; Quake, S. Phys. Rev. Lett. 2001, 86, 4163-4166.

Production of Unilamellar Vesicles

Langmuir, Vol. 19, No. 7, 2003 2877

Figure 9. (A) Phase contrast microscopy image of Span80-egg-PC (80:20) vesicles. The arrow points to a budding vesicle. (B) Confocal microscopy image of a PABuPAM (diblock copolymer) vesicle tagged with 1 mol % NBD-PC. The distribution of the polarized fluorescence intensity confirms that the tagged lipids are incorporated into the polymer bilayer and are ordered normal to surface.

the vesicle solution than in the inverted emulsion. We believe that only a fraction of the large emulsion droplets pass through the interface to form vesicles because most break at the interface. The breakage mechanism remains unclear, but experimental evidence suggests that the surface concentration of lipid at the oil-water interface plays an important role. This monolayer is also affected by the kinetics of interfacial lipid adsorption at the oil-water interface between the two bottom phases. We have observed in our study of asymmetric vesicles13 that several hours of equilibration time are also required to achieve sufficient lipid coverage of the intermediate phase monolayer for optimum conversion of emulsion droplets into vesicles, and that shorter equilibration times lead to an insufficient coverage and uncontrolled lipid composition. Unfortunately, when the interface is left to equilibrate for a much longer period of time, we observe a whitish cloud at the oil-water interface corresponding to an accumulation of lipid in multilayer structures. This very delicate balance between equilibration, which is required for the optimum formation of the vesicles, and spontaneous emulsification, which degrades the formation, creates a significant limitation in the technique. In practice, we allow this interface to equilibrate for more than two, but fewer than 3 h, providing the best compromise choice. The depletion of lipid from the oil-water interface can also be responsible for transfer failure. As the emulsion droplets pass through the interface, they pick up part of the monolayer for bilayer completion. New lipids do not adsorb quickly enough to replenish the interface during the relatively fast sedimentation stage. This suggests that the area of the lower interface should be comparable to the total area of bilayer required for optimum vesicle production, which limits vesicle yield. Applications Here we present some applications that utilize this method of vesicle production; these examples exploit some of the advantages of this technique: the possibility of producing vesicles using a variety of bilayer constituents, the encapsulation efficiency, and the preservation of biological activity. Polymer or Surfactant Constituent Molecules. This technique is not limited to lipids. Just as electroformation has been used to form more resistant polymersomes from diblock copolymers,36 this technique can be adapted for use with a wide variety of amphiphilic molecules to produce polymeric bilayers or mixed bilayers (36) Discher, B. M.; Won, Y. Y.; Ege, D. S.; Lee, J. C. M.; Bates, F. S.; Discher, D. E.; Hammer, D. A. Science 1999, 284, 1143-1146.

where high encapsulation yield is preserved. In fact, a wide selection of amphiphilic molecules can be used as long as they can self-assemble at the oil-water interface and can stabilize an inverted emulsion. We have successfully made vesicles using inverted emulsions prepared with a wide variety of phospholipids both with and without added cholesterol, synthetic surfactants (either cationic or nonionic), and diblock copolymers. Two examples are presented in Figure 9. In Figure 9A, we show a phase contrast microscopy image of vesicles prepared from an emulsion stabilized by 80:20 Span-80:egg-PC and formed by sonication. During the first 15 min following their formation, most of the largest vesicles presenting an excess area fluctuate and bud to form two smaller vesicles. The arrow shows a vesicle about to bud. Since we did not observe such behavior in other surfactants, we believe that the presence of Span80 makes these vesicles more flexible leading to budding. After this reorganization, the vesicles remain stable for several days. When the percentage of Span-80 used is greater than 80%, the vesicles formed are unstable and break into micelles within a matter of hours. An image obtained by confocal microscopy of one of the vesicles made with a diblock-copolymer PABuPAM, Mw ) 16 000 g/mol, tagged with 1 mol % NBD-PC is shown in Figure 9B. Nematic ordering of the molecules in the bilayer is observed in the polarization of the fluorescence; the brighter poles of the vesicle correspond to fluorescent light polarized parallel to the analyzer, and the darker poles correspond to fluorescent light polarized perpendicular to the analyzer, indicating that the fluorescent lipids are inserted in the polymer bilayer normal to the interface as in a lipid bilayer. Although this bipolarity is less pronounced than in a thinner lipid bilayer, it does suggest that the polymer bilayer is fairly ordered. Microreactors. Using this preparation technique, active molecules can be isolated from the bulk by encapsulating them in vesicles. Once encapsulated, macromolecules are unable to diffuse out, but a limited quantity of small molecules can diffuse from the bulk through the bilayer into the vesicles. Thus it is possible to design vesicles for use as microreactors, where the reaction of macromolecules occurs in a controlled, isolated environment. This could have particular application in the encapsulation of biologically important molecules, provided their biological activity can be preserved during vesicle synthesis. To demonstrate the use of vesicles as microreactors, it is necessary to show that the vesicles successfully encapsulate macromolecules, that we can control the transport of initiators across the bilayer, and that we can retain biological activity during the vesicle production process. As an example, we demonstrate the

2878

Langmuir, Vol. 19, No. 7, 2003

Figure 10. Vesicles formed from an emulsion stabilized by egg-PC:cholesterol (80:20), encapsulating Mw ) 10 000 dextran tagged with Texas Red, and produced by extrusion through a 2-µm-diameter filter. In this case, the vesicles are not bound to the substrate, so their intensity varies depending on their position in the focal plane.

polymerization of actin monomers triggered by an increase of the concentration of magnesium. (1) Encapsulation. To demonstrate the encapsulation of macromolecules, we measured the fluorescence of 2-µm vesicles prepared from an egg-PC: cholesterol stabilized emulsion where the aqueous phase contained 1 wt % of a fluorescently labeled dextran (Mw ) 10 000 g/mol), a water-soluble polysaccharide. Using fluorescence microscopy, we observed that the fluorescence originates only in the interior of the vesicles with little background intensity, confirming that the dextran remains encapsulated within the vesicles, as shown in Figure 10. We have also been able to encapsulate dextran with Mw ) 200 000 g/mol. (2) Ion Transport across the Bilayer. To demonstrate controlled ion transport, we encapsulated 10 000 Mw dextran conjugated with a calcium sensitive dye, Oregon Green, in vesicles with and without ionophores.37 The ionophore used was A23187 (Sigma, St Louis); when present in the bilayer, it facilitates the transport of divalent cations across the bilayer. We used fluorescence spectroscopy to quantitatively measure changes in fluorescence intensity as a function of the concentration of calcium in the solution. As a control experiment, we first measured the fluorescence intensity (λexc) 488 nm, λem) 535 nm) of the conjugated dextran in bulk buffer solution before encapsulation as a function of the calcium added. Figure 11A shows the changes in the emission spectrum upon addition of calcium. The solid triangles correspond to the intensity of the emission spectrum for this dye without calcium in solution. After the addition of 5 mM calcium chloride, the intensity increased by a factor of 1.7, as shown by the open triangles. The same polymer solution was then emulsified and encapsulated in vesicles. The vesicles were prepared with egg-PC:DOPS (80:20) with and without ionophores. In the presence of ionophores, calcium can diffuse into the vesicles until the concentration gradient is zero, with a corresponding increase in fluorescence intensity with time. Figure 11B shows the time dependence of the fluorescence intensity obtained for the encapsulated calcium-sensitive dextran. At time t ) 0, we observe the fluorescence of the encapsulated polymer without calcium. The addition of 5 mM calcium to the vesicle suspension resulted in an increase of the fluorescence by a factor of 1.7, identical to that observed for the bulk sample. This confirms that the final concentration of calcium inside the vesicles is the (37) Hackl, W.; Barmann, M.; Sackmann, E. Phys. Rev. Lett. 1998, 80, 1789.

Pautot et al.

Figure 11. (A) Emission spectrum of dextran labeled with Oregon Green 488 bapta-1 at λex ) 488 nm. The solid triangles correspond to the emission spectrum of the dextran in buffer without calcium, and the open triangles show the changes in the emission spectrum after addition of 5 mM CaCl2 in solution and are used to calibrate the fluorescent response for the addition of fixed amount of calcium. (B) Time course of the change in fluorescence intensity at λem ) 535 nm for dextran encapsulated in egg-PC:DOPS vesicles upon addition of 5 mM CaCl2 to the vesicle suspension. Ionophores have been incorporated into the bilayers to allow passive transport of divalent cations across the membrane. The curve shows how the fluorescence intensity changes as calcium ions diffuse into the vesicles.

same as the bulk concentration. In contrast, in a control experiment where vesicles were prepared without ionophores, we did not observe any detectable changes in fluorescence after addition of calcium to the suspension, confirming that ion transport across the membrane can be controlled using ionophores. Furthermore, this also quantitatively confirms our previous result that there is no measurable nonencapsulated dextran in the bulk solution and that dextran has been successfully encapsulated. (3) Biocompatibility. A very important feature of vesicles is their biocompatibility. However, this emulsion-based production technique involves several transfers between fluids, not all of which may be completely compatible with biomaterials. To test the preservation of biological activity within our vesicles, we investigated the aggregation of G-actin to form F-actin. We encapsulated G-actin monomers from acanthamoeba14 in egg-PC vesicles with ionophores incorporated in the bilayer and studied the polymerization kinetics of the actin for two concentrations of encapsulated monomer: 5 µM and 15 µM. To follow the kinetics of polymerization, we added 20% N-(1-pyrenyl)iodoacetamide-labeled G-actin to the solution of actin monomer; the fluorescence of pyrene-labeled actin increases significantly upon polymerization, providing a sensitive probe of the reaction.15 We initiated the polymerization by the addition of salt to reach a final concentration of 50 mM KCl and 2 mM MgCl2. The fluorescence intensity from both the encapsulated actin and a control sample of the same actin in bulk solution during the polymerization process for two actin concentrations is shown in Figure 12. A large difference in absolute intensities is measured between the bulk (symbols) and the encapsulated actin (curves), reflecting the dilution factor; the encapsulated data for both samples have been multiplied by a constant factor of 18.5 to rescale them to the same level as the bulk data. Figure 12 shows that, in all cases, the addition of salt generates a strong increase in the fluorescent signal observed. The results obtained for 15 µM actin in bulk

Production of Unilamellar Vesicles

Figure 12. Fluorescence intensity of pyrene-labeled actin as a function of time after adding 50 mM KCl and 2 mM MgCl2 to initiate polymerization. The symbols represent the kinetics of bulk actin at concentrations of 5 µM (empty circles) and 15 µM (plusses), while the curves are the kinetics of actin encapsulated in egg-PC vesicles at concentrations of 5 µM (solid line) and 15 µM (dotted line). The encapsulated actin data have been scaled by a constant consistent with the dilution factor.

(crosses) and the corresponding encapsulated actin (dotted line) overlap, confirming that the time dependence of the polymerization kinetics is identical. Similarly, the curve for 5 µM actin in bulk (circles) overlaps with the corresponding encapsulated actin (solid line). These results confirm that the same polymerization reaction takes place within the vesicle as in bulk and that the concentration and protein activity are both preserved by the encapsulation technique. The results also allow a measurement of encapsulation efficiency. The emulsion droplets that contain the actin have a total volume of 0.165 mL and are transferred into 3.0 mL of final aqueous phase. This results in a dilution factor of 18.2. Thus, for a 100% efficient process of encapsulation and transfer through the interface, the fluorescence of the vesicle suspension is expected to be 18.2 times smaller than the bulk sample. The measured scaling factor of 18.5 corresponds to 98% encapsulation efficiency. Conclusions The lipid behavior at oil-water interfaces makes it difficult to control the emulsification process and reduces the domain where this emulsion-based vesicle production technique produces optimal results. Some of these limitations can be mitigated by consideration of the phase behavior of the lipid/oil/water system and the kinetics of lipid adsorption. Once a protocol is designed that takes these limitations into account, the inverted emulsion

Langmuir, Vol. 19, No. 7, 2003 2879

technique provides a way to simultaneously improve the encapsulation yield and to control the bilayer composition, while producing unilamellar lipid vesicles. We have shown that this technique can be used to prepare surfactant vesicles or polymersomes; in fact, the use of conventional surfactants or diblock copolymer improves our control of the emulsification process. We can encapsulate a wide variety of macromolecules at any desired concentration while preserving physiological conditions with encapsulation efficiencies of up to 98%. This technique is one of the very few means of conveniently encapsulating large macromolecules with radii of gyration as large as 1 µm, as may be required for potential application such as drug delivery and gene therapy. Moreover, these carriers can be “functionalized” using biotin-conjugated lipids, providing a primary binding site for streptavidin. These vesicles can then react with the desired biotinylated antibodies and provide a convenient way to design immuno-liposomes,38-40 which can specifically bind to cellular receptors. Furthermore, we have demonstrated that this technique is compatible with biological activity and that the vesicles produced with this method can be used as micro-bioreactors. This provides the possibility of controlling reactions in a confined space. This method also offers a new way to design chemical sensors where vesicles could be used to locally probe the ion concentration in solution as well as multiencapsulation devices with vesicles containing subcompartments filled with different molecules. In summary, emulsion-based vesicle production provides a new means of making vesicles that may find uses in a wide range of materials science. Acknowledgment. We thank E.M. Ostap for assistance in purifying G-actin and for use of his laboratory for the actin experiments and G. Whitesides for use of his spectrofluorimeter. We are grateful to J. Crocker, D. Discher, and E. Weeks for their useful suggestions. S.P. is particularly grateful to C. Gay and M.T. Valentine for help in the preparation of this article and to P. Poulin for his encouragement at the early stage of this work and for discussions as the project progressed. We also thank Rhodia (Cranburry, NJ) for supplying the diblock copolymers. This work was supported by Rhodia, Kraft Foods, and by the NSF (DMR-9971432). LA026100V (38) Szebeni, J. Crit. Rev. Therapeutic Drug Carrier Systems 1998, 15, 57-88. (39) Park, J. W.; Hong, K.; Kirpotin, D. B.; Papahadjopoulos, D.; Benz, C. C. Immunoliposomes for Cancer Treatment; Academic Press: New York, 1997; Vol. 40. (40) Harokopakis, E.; Hajishengallis, G.; Michalek, S. M. Infect. Immun. 1998, 66, 4299-4304.